dust is a new France-based AI startup focused on improving team productivity by breaking down internal silos, revealing vital knowledge, and providing tools to build custom internal applications. At its core, Dust uses large language models (LLMs) on internal company data to give team members new superpowers.
The company was co-founded by Gabriel Hubert and Stanislas Polu, who have known each other for over a decade. Their first startup, Totems, was acquired by Stripe in 2015. They both worked at Stripe for a few years thereafter before parting ways.
Stanislas Polu joins OpenAI, where he spent three years researching reasoning for LLMs, while Gabriel Hubert becomes Head of Product allen.
They teamed up again to create Dust. Unlike many AI startups, Dust isn’t focused on creating new large language models. Instead, the company wants to build applications on top of LL.M.s developed by OpenAI, Cohere, AI21, and others.
The team first worked on platform Can be used to design and deploy large language model applications. It then focused its efforts on one use case in particular – centralizing and indexing internal data so that the LL.M. could use it.
From internal ChatGPT to next-gen software
There are connectors that constantly fetch internal data from Notion, Slack, Github, and Google Drive. This data is then indexed and made available for semantic search queries. When a user wants to perform some action using a Dust-powered application, Dust finds the relevant internal data, uses it as context for the LLM, and returns the answer.
For example, suppose you have just joined a company and are working on a project that was started a while ago. If your company promotes transparency in communications, you’ll want to look for information in existing internal data. But the internal knowledge base may not be up to date. Or it might be hard to find the reason for doing so, since it’s already been discussed in an archived Slack channel.
Dust isn’t just a better internal search tool because it doesn’t just return search results. It looks up information across multiple data sources and formats answers in a way that’s more useful to you. It can be used as a kind of internal ChatGPT, but it can also be used as the basis for new internal tools.
“We believe natural language interfaces are going to disrupt software,” Gabriel Hubert told me. “Five years from now, it would be disappointing if you still have to click edit, settings, preferences to decide that your software should behave differently. We see our software adapt more and more to your individual demand, because that’s how you are, and because that’s how your team is – because that’s how your company is.”
The company is working with design partners to implement and package the Dust platform in a variety of ways. “We think there are a lot of different products that can be created in the domain of enterprise data, knowledge workers and models to support them,” Stanislas Polu told me.
It’s early days for Dust, but the startup is exploring an interesting problem. We face many challenges when it comes to data retention, hallucinations, and all that LL.M. brings with it. Perhaps as LL.M. develops, hallucinations will become less of an issue. Perhaps for data privacy reasons, Dust will eventually create his own LL.M.
Dust has raised $5.5 million (€5 million) in a seed round led by Sequoia Capital, with participation from XYZ, GG1, Seedcamp, Connect, Motier Ventures, Tiny Supercomputer, AI Grant, and numerous angel investors, such as from Olivier Pomel of Datadog, Julien Codorniou, Julien Chaumond of Hugging Face, Mathilde Colin of Front, Charles Gorintin and Jean-Charles Samuelian-Werve of Alan, Eléonore Crespo and Romain Niccoli of Pigment, Nicolas Brusson of BlaBlaCar, Howie Liu of Airtable, PhotoRoom Mathieu Rouiff, Igor Babuschkin and Irvan Bello.
Taking a step back, Daxter believes the LL.M. will dramatically change the way companies operate. Products like Dust work better in a company that promotes radical transparency over information retention, written communication over endless meetings, and autonomy over top-down management.
If LLMs deliver on their promise and dramatically increase productivity, some companies will gain an unfair advantage by adopting these values, as Dust will unlock much untapped potential for knowledge workers.